Using Self-Organizing Maps to Map Market Regimes

Summary

This article explains how a Kohonen self-organizing map (SOM) can organize multidimensional market features into a two-dimensional grid. Each input pattern is matched to its closest neuron using Euclidean distance, then the winning neuron and nearby neurons adjust their weights toward that pattern. The neighborhood update helps preserve similarity relationships on the map, which can make clusters and transitions among market states easier to inspect without labeled training classes.

The article outlines a 20-by-20 map with 400 input features and describes integrating SOM activation into an MQL5 trading model. It presents code-level mechanics and reports a test whose Sharpe ratio nearly reached 4, but gives no detailed test setup or supporting statistics in the supplied text. The implementation is framed as a research platform rather than a finished strategy. Its claims about noise robustness and likely movement between mapped states are not substantiated with rigorous comparative evidence; historical validation, parameter selection, risk controls, and live evaluation remain necessary.

Key ideas

• A SOM assigns each market feature vector to the neuron whose weights are closest by Euclidean distance.
• The winning neuron and its map neighbors move toward the input according to a Gaussian neighborhood function.
• The resulting grid can visualize clusters of similar market conditions without labeled regime classes.
• The article describes incorporating SOM activations into an MQL5 trading model, but offers limited detail about evaluation.
• The reported Sharpe result lacks a test protocol in the supplied text and does not establish live performance.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.